JP5520380B2 - Methods, products, and apparatus for creating and visualizing temperature zones - Google Patents

Description

  The present invention relates to airflow analysis, and more particularly to techniques for using airflow distribution to model a thermal zone in a space such as a data center.

  For large computing facilities (or data centers), the energy caused by rising energy costs, energy demand and supply, and the proliferation of information and communications technology (ICT) equipment that demands high performance Consumption has become an important issue. Data centers consume approximately 2% of total global power, or 183 billion kilowatt hours (KW), and this power consumption is growing at a rate of 12% annually. A significant percentage of this power consumption, ie up to 50%, is devoted to cooling the heat generating equipment. Therefore, improving the energy and cooling efficiency of the data center is very important. Although the best practice is widely published, data center operators are struggling to set an appropriate amount of cooling. In particular, it is difficult to consider the different heat densities in the data center (ie, different regions in the data center may require very different amounts of cooling).

US Pat. No. 7,366,632 US patent specification 11/750325

"Uncovering Energy-Efficiency Opportunities in Data Centers" by Hamann et al., IBM Journal of Research and Development, vol.53, no.3 (2009) “Fluid Mechanics” by L.D.Landau, PergamonPress (1959) “The Finite Element Method: Linear Static and Dynamic Finite Element Analysis” by T.J.R.Hughes, Chapter 1, Dover Publications (2000) "Remarks Around 50 Lines of Matlab: ShortFinite Element Implementation", Numeric Algorithms 20, 117-138 (1999) by J. Alberty et al. "Methods and Techniques for Measuring and Improving Data Center Best Practices" by H.F.Hamann et al., IEEE Proceedings of ITherm 2008 Conference, Orlando, Florida, pages 1146 to 1152 (May 2008)

  Therefore, techniques that are aimed at identifying the heat density in the data center and thereby increasing the cooling efficiency are desirable.

The present invention provides a technique for using airflow distribution to model temperature zones. In one aspect of the invention, a method is provided for modeling a temperature zone in a space, such as a data center. The method includes the following steps. A graphic representation of the space is provided. At least one region (hereinafter also referred to as a domain ) is defined in the space for modeling. A mesh is created in a domain by subdividing the domain into a set of subdivided regions (hereinafter also referred to as subdomains ) interconnecting a plurality of nodes. Where the airflow in the domain enters the domain (airflow inflow place, hereinafter also referred to as airflow source ) and where the airflow exits the domain (airflow outflow place, hereinafter also referred to as airflow sink ) Identified. Airflow measurements are obtained from one or more of the airflow sources and sinks. Airflow measurements taken from the airflow source and sink are used to determine the airflow velocity vector at the center of each subdomain. Each velocity vector is traced to one of the air flow sources, and the combination of traces to a given one of the air flow sources represents a temperature zone in the space.

  The invention and other features and advantages of the invention will be more fully understood by reference to the following detailed description and drawings.

FIG. 2 illustrates an exemplary data center according to an embodiment of the present invention. FIG. 6 illustrates an exemplary method for modeling a temperature zone according to an embodiment of the present invention. FIG. 3 illustrates an exemplary finite element mesh according to an embodiment of the present invention. FIG. 4 is a graph illustrating a data center according to an embodiment of the present invention. FIG. 5 is an enlarged view showing a part of the graph of FIG. 4 showing a mesh according to an embodiment of the present invention. FIG. 5 is an enlarged view showing a portion of the graph of FIG. 4 showing airflow potential and airflow velocity vectors in accordance with an embodiment of the present invention. FIG. 5 is an enlarged view showing a portion of the graph of FIG. 4 showing traces and corresponding temperature zones, in accordance with an embodiment of the present invention. FIG. 3 illustrates an exemplary graphical interface including a data center diagram in accordance with an embodiment of the present invention. FIG. 9 illustrates a method for editing the graphical interface of FIG. 8 to reflect changes in a data center according to an embodiment of the present invention. FIG. 9 illustrates temperature zone visualization on the graphical interface of FIG. 8 in accordance with an embodiment of the present invention. FIG. 6 is a schematic diagram illustrating cooling capacity in a data center according to an embodiment of the present invention. FIG. 6 illustrates an exemplary method for tracing an air flow velocity vector to an air flow source, in accordance with an embodiment of the present invention. FIG. 3 illustrates an exemplary apparatus for modeling a temperature zone in a space according to an embodiment of the present invention.

  Provided herein are techniques for dynamically creating and visualizing temperature zones that allow for a better provision and more efficient use of cooling for a data center. Although the technique is described in the context of a data center cooling system, it should be noted that the concepts presented herein are generally applicable to cooling systems and / or heating systems in general. I want to be.

FIG. 1 is a diagram illustrating an exemplary data center 100. The data center 100 captures hot air (typically from the top through one or more return airs in an air conditioning unit ( ACU ) ) and cool air from the underlying floor plenum ( plenum) with a raised floor cooling system 102 (sometimes also referred to as a computer room air conditioner (CRAC)) with an air conditioning unit (ACU) exhausting. Hot airflow through the data center 100 is indicated by white arrows 110 and cooling air through the data center 100 is indicated by black arrows 112. In the following description, the data center above the ground floor plenum may simply be referred to as a high floor, and the ground floor plenum may simply be referred to as a plenum. Thus, by way of example only, as shown in FIG. 1, the ACU draws warm air from the elevated floor and releases cooling air into the plenum (see below).

  In FIG. 1, server rack 101 uses front-to-back cooling and is placed on an elevated floor 106 with a base floor 104 underneath. That is, according to this method, the cooling air is drawn through the front surface (intake port) of each rack, and the warm air is discharged from the back surface (exhaust port) of each rack. Cooling air drawn into the front of the rack is supplied to the intake of each internal information technology (IT) equipment component (eg, server). The space between the raised floor 106 and the ground floor 104 defines a ground floor plenum 108. The underfloor plenum 108 serves as a conduit for transferring cooling air from the ACU 102 to the rack, for example. As shown in FIG. 1, the raised floor 106 consists of a plurality of floor tiles, some of which are perforated. In a properly organized data center (such as the data center 100), the rack 101 is configured to have a warm air path and a cold air path configuration, i.e., having inlets and outlets in alternating directions. That is, the cooling air is blown into the raised floor 106 from the base floor plenum 108 into the cold air passage through the perforated floor tiles 114 (also referred to as vents). Thereafter, the cooling air is drawn into the rack 101 through the intake port on the rack intake port side, and discharged into the warm air passage through the exhaust port on the rack exhaust port side.

  The ACU typically receives chilled water from a refrigeration plant (not shown). Each ACU typically includes a blower motor for circulating air through the ACU, for example, to blow cooling air into the ground floor plenum. Thus, in most data centers, the ACU is a simple heat exchanger that consumes mainly the power required to blow cooling air into the ground floor plenum. There are typically one or more power distribution units (PDUs) (not shown) that distribute power to the server rack 101.

  It is important to optimize the efficiency of the ACU. For example, see “Uncovering Energy-Efficiency Opportunities in Data Centers” by Hamann et al., IBM Journal of Research and Development, vol. 53, no. 3 (2009) (hereinafter referred to as “Hamann”). To that end, it is useful to consider the ACU utilization level (ie utilization (UT) = removed heat / nominal heat load removal capability) or performance factor (COP) (COP = removed heat / power consumption for ACU fan). is there. The utilization level of ACU in some data centers has been found to be as low as 10% (COP˜1.8). However, when efficiency optimization is practiced, the utilization level (even with some redundancy) is potentially in the range of about 80% to about 100%, and the corresponding COP is from about 14 each. It can be about 18. Most data centers require some redundancy. For example, in a data center with eight ACUs, redundancy allows for one ACU failure (eg, due to a mechanical failure). Therefore, basically, there is no N + 1 redundancy (N = 8), so the target utilization level cannot be 7/8 (87.5%).

  One of the obstacles to optimizing ACU usage is the lack of visibility, i.e., the inability to recognize which physical areas or zones in the data center are supplying cooling air to different ACUs. A temperature zone is a physical region (2D (2D)) or volume (3D (3D)) within a data center. Each ACU, sometimes referred to herein as a “supply zone”, supplies air to a particular temperature zone within the data center, typically applied to the plenum. Each ACU is sometimes referred to herein as a “return zone” and also obtains return air from a particular temperature zone within the data center, typically applied to elevated floors.

  It is not uncommon for a data center to have more than 50 ACUs that are distributed somewhat randomly throughout the space of the data center. The temperature zone (supply zone and / or return zone) arising from each ACU depends not only on the airflow generated by each ACU, but also on the placement of vents or perforated tiles throughout the data center (see below). As such, the vents or perforated tiles are at least partially directed to where the air from the ACU is directed in the data center). It is not uncommon for a data center to have more than 1000 vents or perforated tiles. Since these temperature zones are based on the actual airflow contribution of each ACU, a corresponding efficiency or coefficient of performance (COP) can be assigned to each temperature zone of each individual ACU. The airflow distribution throughout the data center depends on many aspects. The technique uses the simplest form of air flow distribution in the data center by applying the zone concept to the plenum. The ACU uses a fan to exhaust air into the plenum, with the result that the plenum is pressurized. The placement of vents or perforated tiles will determine where air will escape the plenum on the raised floor. This vent / perforated tile arrangement determines the zone (ie, which area is supplied by which ACU).

  Disclosed herein are techniques for modeling, ie creating and visualizing, these temperature zones. As will be described in more detail below, a velocity field is used to define or create a temperature zone (ie, the zone must be created because it is not predetermined). Zone creation / definition is an important aspect of the present technique because it provides a method for determining (by COP measurement) how efficiently each ACU is being used. FIG. 2 is a diagram illustrating an exemplary method 200 for modeling temperature zones in a room, for example, a space such as data center 100. In step 202, a graphical representation of the space is provided. For example, in FIG. 4 described below, a two-dimensional graphical representation of the data center is provided. In the case of a data center, the graphical display can include a layout of, for example, server racks, vents or perforated tiles, and ACUs. A graphical display can be created using a software application that “draws” an asset, such as a server rack. One example is a mobility management technology software application (used to create the graphical display shown in FIG. 4).

  In step 204, at least one domain for temperature zone modeling is defined in the space. Each domain can be a two-dimensional or three-dimensional domain. As described below, in an exemplary embodiment in which the data center is modeled, the domain is defined by the dimensions of the ground floor plenum. As described further below, multiple domains can be defined for a given space. Depending on the particular application, such as the physical layout of the space to which the technique is applied, the domain can include the entire space, or a portion thereof.

In step 206, because the finite element is employed to model the space, each domain, by subdividing the set of interconnects to salicylate subdomains multiple nodes, finite element on each of the domain A mesh is created. As described in detail below, subdomains (also referred to herein as “elements”) can be triangular (in the case of a two-dimensional domain) or tetrahedron (in the case of a three-dimensional domain). The use of triangles and tetrahedrons is a standard choice in finite element methods. Nodes correspond to x and y coordinates (for a two-dimensional domain) or x, y, and z coordinates (for a three-dimensional domain).

  In step 208, an airflow source (where the airflow enters the domain) and an airflow sink (where the airflow exits the domain) are identified in the domain. Looking at the data center, if the domain includes an underfloor plenum, the perforated tiles can be considered as airflow sinks (cooling air supplied by the ACU exits the underfloor plenum and enters the elevated floor. The ACU can be considered an airflow source (because it is in the perforated tiles of the site) (because the airflow originates from the ACU / enters the ground floor plenum). On the other hand, if the domain includes elevated floors, the perforated tiles can be considered as an air flow source (because the cooling air from the plenum is in the perforated tiles where it enters the elevated floor) and the ACU can be considered as a sink (Because the warm air circulates back into the ACU and exits the elevated floor with the ACU).

  In step 210, airflow measurements are obtained from one or more of the airflow sources and sinks. According to an exemplary embodiment, these airflow measurements are obtained using a moving measurement technique (MMT). MMT is described in US Pat. No. 7,366,632 (hereinafter referred to as “US Pat. No. 7,366,632”) entitled “Method and Apparatus for Three-Dimensional Measurements” issued to Hamann et al. MMT V1.0 is a technology for optimizing data center infrastructure to improve energy and space efficiency, and for a given space and most efficient energy utilization, a given Advanced metrology techniques for high-speed data center measurement / investigation (see, eg, US Pat. No. 7,366,632) to optimize the data center within the temperature envelope, and metric-based assessment and Implementation of a database-based best-practice (for example, US Patent Specification No. 11/750325 named “Techniques for Analyzing Data Center Energy Utilization Practices”, designated as reference number YOR920070242US1, filed by Claassen et al. Combination with reference).

  In step 212, airflow measurements obtained from the airflow source and sink are used to determine the airflow velocity vector at the center of each subdomain (element). An exemplary process for determining these airflow velocity vectors using latent flow theory is described in detail below.

  In step 214, each velocity vector is traced to one of the air flow sources. Looking at a data center model where the domain has a ground floor plenum, each velocity vector can be traced to a specific ACU. An exemplary process for tracing the velocity vector is described in detail below. The combination of traces to a given one of the airflow sources (eg ACU) represents a temperature zone within the space. An exemplary temperature zone is shown, for example, in FIG. 10 described below.

  As shown in FIG. 2, steps 208-214 of method 200 are removed from one or more of the sources / sinks, one or more new sources / sinks added to the domain, or from the domain. Iterating in response to changes in airflow in one or more of the sources / sinks, or all of them. For example, referring to the data center, air flow may increase or decrease at one or more of the vents or perforated tiles, or one or more of the vents or perforated tiles. May be rearranged or removed, or both. As will be explained in detail below, advantageously, these elements can be reused in subsequent iterations of the method as long as the domain / finite element mesh does not change.

  As the above steps are repeated, updated air flow measurements (if available) can be obtained from the air flow source and sink, so instant techniques are in space. Be sensitive to changes in the situation. Merely by way of example, steps 210-214 may be repeated periodically (eg, at predetermined time intervals) to obtain updated airflow measurement data. The predetermined time interval can be set based on, for example, the frequency at which updated measurement data is available, the frequency at which changes occur in the space, or both.

  In step 216, the cooling capacity is determined for each airflow source. For example, referring to data center modeling, the cooling capacity of each vent or perforated tile can be determined. An exemplary process for determining cooling capacity is described in detail below.

  As is apparent above, in an exemplary embodiment, a potential flow theory (assuming that the air density is constant (independent of temperature), proceeds beyond the boundary, and that the viscous force is negligible ( potential flow theory). For a general explanation of latent flow theory, see, for example, “Fluid Mechanics” by L.D.Landau et al., PergamonPress (1959).

ここで、ｆは流れのソースまたはシンクを表し、ｖ_ｘ、ｖ_ｙ、およびｖ_ｚは、それぞれｘ、ｙ、およびｚ方向の速度成分である。本技法に従って、乱流モデル（turbulence model）および散逸（dissipation）を含むことが可能な、他の複数の包括的な偏微分方程式（ＰＤＥ）が使用可能であることに留意されたい。 If the air velocity v (v _x , v _y , v _z ) is assumed to be non-rotational, that is, curl v = 0, the velocity is assumed to be the slope of the scalar function φ. be able to. This function φ is called “velocity (or air amount) potential” and satisfies the Poisson equation. In other words, the (air) velocity field corresponds to a solution of the following equation with appropriate boundary conditions.

Where f represents a flow source or sink, and v _x , v _y , and v _z are velocity components in the x, y, and z directions, respectively. Note that a number of other comprehensive partial differential equations (PDEs) can be used in accordance with the present technique, which can include turbulence models and dissipation.

＝−（測定された）穿孔タイルからの出力速度）として、ＡＣＵへの戻りを「シンク」（φ＝０）として、モデル化することが可能であるが、同時にラックは、シンク（

＝（測定された）吸気口ラック流れ）およびソース（

＝−（測定された）排気口ラック流れ）である。前述のように、ソースおよびシンクはドメインに依存して変化する可能性がある（たとえば、ドメインが下地床プレナムを含む場合、穿孔タイルおよびＡＣＵはそれぞれシンクおよびソースであるとみなすことが可能であり、他方で、ドメインが高床を含む場合、穿孔タイルおよびＡＣＵはそれぞれソースおよびシンクであるとみなすことが可能である）。他の代替は、数式１における非ゼロの右側ｆを介して、ソースおよびシンクをモデル化することである。ソースまたはシンクがＰＤＥの解ドメインの境界上に配置される場合、前者の手法が適切である（ノイマン境界条件の指定に対応する場合）。解ドメイン内に配置された（すなわち境界上にない）ソースまたはシンクの場合、数式１における非ゼロの右側ｆを介したモデル化の方が好適となる。 To provide boundary conditions for the aforementioned problems, for example, vents or perforated tiles (or the output of an ACU) can be “sourced” (

=-(Measured) output speed from perforated tiles) can be modeled as a “sink” (φ = 0) return to the ACU, but at the same time the rack is

= (Measured) inlet rack flow) and source (

=-(Measured) outlet rack flow). As described above, the source and sink may vary depending on the domain (eg, if the domain includes a ground floor plenum, the perforated tiles and the ACU may be considered to be sinks and sources, respectively. On the other hand, if the domain includes raised floors, the perforated tiles and the ACU can be considered as sources and sinks, respectively). Another alternative is to model the source and sink via the non-zero right f in Equation 1. If the source or sink is located on the boundary of the PDE solution domain, the former approach is appropriate (if it corresponds to the specification of the Neumann boundary condition). For sources or sinks located in the solution domain (ie not on the boundary), modeling via the non-zero right f in Equation 1 is preferred.

  Herein, a finite element solver is implemented to calculate the airflow potential. For a standard reference on finite element methods, see TJRHughes' “The Finite Element Method: Linear Static and Dynamic Finite Element Analysis”, Chapter 1, Dover Publications (2000) (first edition published in 1987 by Prentice-Hall). Please refer. One embodiment of the technique is implemented in the C programming language, after which some similar finite element implementations are described in J. Alberty et al., “Remarks Around 50 Lines of Matlab: ShortFinite Element Implementation”, Numerical Algorithms 20, 117- 138 pages (1999) (hereinafter “Alberty”), implemented in Matlab®.

  This solution requires a mesh specification consisting of a set of nodes in the specified domain and the triangles (in the case of a two-dimensional domain) or tetrahedron (in the case of a three-dimensional domain) that connect these nodes. . As is apparent above, the triangle (or tetrahedron) is also referred to herein as a mesh element. The node corresponds to the (x, y) coordinates of the two-dimensional domain (or (x, y, z) coordinates of the three-dimensional domain). An example of a finite element mesh is shown in FIG. That is, FIG. 3 is a diagram illustrating an exemplary finite element mesh 300 that includes a set of nine nodes and eight elements in a two-dimensional domain. A legend for each of the eight elements E and corresponding nodes is provided on the right side of the mesh.

が、Ｎ個の基底関数Ψ_ｉの線形結合、ｉ＝１、・・・、Ｎとして求められる。

がメッシュ内のｊ番目のノードを示すものとすると、ｉ＝ｊの場合は

そうでなければ

であるように、基底関数が選択される。これは、有限要素近似の典型であり、数式３の線形結合内の係数

は、ｉ番目のノードの解の近似値にも対応する。Albertyにあるように、区分線形（piecewise linear）基底関数を使用して、数式１の解を近似することができる。ここでも、これは有限要素近似における標準的な選択である。ガラーキン有限要素離散化を好適な境界条件で数式１に適用すると、以下の連立一次方程式が得られる。

数式４の連立方程式の結果を求めるために、ガラーキン有限要素離散化を数式１に適用するためのプロセスは、当業者であれば明らかであるため、本明細書では詳細に説明しない。数式４の解

は、有限要素メッシュのノードでの空気流ポテンシャルφへの近似解を与える。問題は、数式２で速度場を定義する場合、ポテンシャルの勾配を得ることにあるため、数式４の線形システムが解かれると、数式３と数式４のシステムの解

の線形結合から、勾配φへの数値近似を得ることができる。これにより、数式２で定義されたような空気流場が提供される。しかしながら基底関数の選択により、メッシュ・ノードでの勾配に対する近似を得ることは意味がない。代わりに、各要素内の点での近似が有効である。各要素の中心は便宜的かつ標準の座標点、すなわち、速度場を近似するための場所として選択することができる。次に、以下で説明するように、空気流の流路（trajectories）から温度ゾーンを定義することができる。 Approximate solution for Equation 1

Is obtained as a linear combination of N basis functions ψ _i , i = 1,.

Let j denote the jth node in the mesh, if i = j

Otherwise

The basis function is selected such that This is typical of the finite element approximation, and the coefficient in the linear combination of Equation 3

Corresponds to the approximate value of the solution of the i-th node. As in Alberty, a piecewise linear basis function can be used to approximate the solution of Equation 1. Again, this is the standard choice in finite element approximation. When Galerkin finite element discretization is applied to Equation 1 with suitable boundary conditions, the following simultaneous linear equations are obtained.

The process for applying Galerkin finite element discretization to Equation 1 to determine the results of the simultaneous equations of Equation 4 will be apparent to those skilled in the art and will not be described in detail herein. Solution of Equation 4

Gives an approximate solution to the airflow potential φ at a node of a finite element mesh. The problem is that when the velocity field is defined by Equation 2, the potential gradient is obtained. Therefore, when the linear system of Equation 4 is solved, the solution of Equation 3 and Equation 4 is solved.

From this linear combination, a numerical approximation to the gradient φ can be obtained. This provides an air flow field as defined by Equation 2. However, it does not make sense to get an approximation to the gradient at the mesh node by selecting the basis function. Instead, approximation at points within each element is useful. The center of each element can be chosen as a convenient and standard coordinate point, i.e. a place to approximate the velocity field. Next, as described below, a temperature zone can be defined from the trajectories of the air flow.

  Next, a two-dimensional implementation of the above technique for calculating the airflow potential for a large (eg, greater than 50,000 square feet) data center underfloor plenum is described. In the following, the focus is on the temperature zone on the “plenum”, but the same principle can be applied to “above the plenum” or the entire elevated data center. FIG. 4 is a diagram illustrating a data center graphical display 400 including a plurality of ACUs 402 and perforated tiles 404. Such a graphic display can be created, for example, using an MMT client as follows. Certain areas in the data center, such as area 406, are excluded from the calculation because, for example, area 406 is separated from the plenum. Domain boundary 408 defines the physical dimensions of the entire plenum.

  As described above, once the domain is defined by the ground floor plenum, the ACU can be an air flow source and the perforated tile can be an air flow sink, as applied as a Neumann boundary condition. In this particular embodiment, the right side of Equation 1 is set to zero. As is apparent above, airflow measurements for ACU and perforated tiles can be obtained using MMT. Suitable techniques for obtaining the necessary input data for both ACU and perforated tiles are described, for example, in “Methods and Techniques for Measuring and Improving Data Center Best Practices” by HF Hamann et al., IEEE Proceedings of ITherm 2008 Conference, Orlando, Florida, 1146- Also described on page 1152 (May 2008).

  A magnified view 500 of a portion of the graphical display 400 shown in FIG. 5 shows that a mesh has been created. In this two-dimensional model, nodes are connected to triangles. Sources (eg ACU) and sinks (eg perforated tiles) are labeled and displayed in FIG.

  FIG. 6 is a diagram showing the air flow potential φ and the corresponding air flow velocity vector obtained from the execution of the above-described calculation. The same portion 500 of the graphical display 400 is shown in FIG. Dark areas correspond to high airflow potentials, and light shaded areas correspond to low airflow potentials. As shown in FIG. 6, no boundaries are applied because some ACUs are turned off. The underlying grid refers to a tile grid that is typically used in a data center.

Once the airflow velocity field is calculated (in accordance with Equation 2 above), the airflow from / to each region of the data center is traced back to the origin / return ACU (the airflow velocity vector is Collectively showing airflow streamlines (hereinafter also referred to as airflow patterns), which is why velocity fields are used to define temperature zones (given by these airflow patterns)). An exemplary method for tracing airflow within a data center is described in connection with the description of FIG. 12 below, for example. In this particular example, airflow is traced across the underlying floor plenum, but can be performed in the same manner for elevated floors. A trace connects a particular ACU and its corresponding temperature zone.

These traces (also referred to herein as “air flow channels”), as well as the corresponding temperature zones, are the paths that the individual particles follow. Again, the same portion 500 of the graphical display 400 is shown in FIG. In FIG. 7, an overview of temperature zones is shown (by zone boundaries). The ACU of the shaded pattern (hatched pattern) is off. If conditions change, such as removing perforated tiles, placing different tile types, or turning off the ACU, repeat the calculation and change the data center plenum zone accordingly can do. Different types of tiles have different levels of perforations, which is how you can control how much airflow is forced through the tile for a given pressure gradient. In general, reference is made to the “level” of the opening of the tile. Thus, for example, there are 25% up to 50% open tiles. A 25% open tile is more resistant than a 50% open tile. Resistance or impedance R is a characteristic of the openness of the tile and some other factor. If the pressure difference p is known, the air flow f can be calculated as follows:
p = R × f ²

  In this regard, it should be noted that unless such changes involve modifications to the finite element mesh, nodes and elements do not need to be regenerated each time, thus reducing the time required to repeat the computation. Furthermore, if a direct solution for the linear system is used to solve Equation 4, computation time can be saved as long as the matrix A in Equation 4 remains unchanged (this is because the condition change is source code) This is because if the result corresponds only to the correction in the flow measured at the sink, this result corrects only the right side b of Equation 4). Typically, the direct solution employs two phases: the numerical factorization of the coefficient matrix followed by the solution of the system where the matrix is factored, so that as long as the matrix remains unchanged, the coefficients It is only necessary to perform the numerical factorization of the matrix once. Since numerical factorization saves the most time in two phases, performing factorization only when strictly necessary results in significant computational time savings.

The technique can include the development of superposition principles that can provide a way to resolve each airflow pattern with varying conditions faster by avoiding redundant computations. Specifically, Equation 1 for solving for potential φ can be rewritten as ∇ · (∇φ) = f, where ∇ · is a divergence operator. Using the superposition principle, the two solutions φ ₁ and φ ₂ (or in general any number of solutions) of Equation 1 are summed, and φ ₁ and φ ₂ are the solutions to Equation 1 for the same domain (geometry) As long as there is a third solution φ ₃ = φ ₁ + φ ₂ can be obtained. That is, if φ ₁ solves ∇ · (∇φ ₁ ) = f ₁ and φ ₂ solves ∇ · (∇φ ₂ ) = f ₂ , then φ ₃ = φ ₁ + φ ₂ becomes ∇ · ( Resolve ∇φ ₃ ) = f ₁ + f ₂ For example, φ ₁ is a solution obtained with only ACU _{1 on} (at a given fan speed setting) and all other ACUs off, and φ ₂ is only ACU ₂ (with a given fan). Suppose the solution is on (at the speed setting) and is obtained with all other ACUs off. ACU1 only velocity field in the scenario state ON _{v 1} = a ∇φ _1, velocity field only ACU2 is turned on scenario is _{v 2} = ∇φ _2. Then v ₃ = v ₁ + v ₂ corresponds to the velocity field in a scenario where ACU1 and ACU2 are on and all other ACUs are off.

  In an exemplary embodiment, graphics software (eg, an MMT client) is used to provide a graphical display, eg, a data center, define domains, edit sinks and sources, and define sinks and sources Send sensor data (if available) for initialization, initialize calculations, post-process and visualize temperature zones. The MMT client is a software application that allows a graphic display of the data center layout and visualization of airflow channels, zones, and airflow vectors.

  See, for example, FIG. 8, which illustrates an exemplary graphical interface 800 (eg, as shown on a video display such as a computer monitor, shown below), including, for example, a data center graphic display 801. To do. For illustrative purposes, structures in the data center such as server racks and perforated tiles are labeled in the graphical display 801 (and may or may not be labeled in the actual implementation of the process). Have sex). The plenum domain is defined as a polygon of x and y points, for example, by a boundary 806. Here, two domains, outer and inner, are defined (the inner domain is not visible in this depiction). The outer domain determines the perimeter of the data center, and the inner domain covers a separate control room. Dirichlet boundary conditions are also defined (potential is set to a constant value, most often zero).

ρは空気密度、Ａはタイルの面積、Ｋは損失係数である。圧力差と空気流とは以下のように関係付けられる。
Δρ＝Ｒ・ｆ^２ _{ａｉｒｆｌｏｗ}
圧力差はリアルタイム・センサで再測定され、Δρ＝Ｒ・ｆ^２ _{ａｉｒｆｌｏｗ}を使用して各タイルを通る空気流を計算する。前述のように、空気流は境界として適用される。 As is apparent above, the air flow at one or more of the sources / sinks can vary. In addition, sources / sinks can be added to the domain and / or removed from the domain. FIG. 9 is a diagram illustrating a method for graphically editing the interface 800 (of FIG. 8) to reflect these changes in the domain. In this particular example, the air flow for the perforated tile is specified. Perforated tiles can move and overlap interactively, and tile characteristics (type, airflow impedance, etc.) can be easily manipulated. In one exemplary embodiment, model inputs can be updated using, for example, real-time sensors located within the plenum. Specifically, by measuring the pressure in the plenum and combining it with the air flow impedance of the vent or perforated tile, the air flow can be automatically calculated and applied to the zoning model. The following formula gives the perforated airflow tile impedance R (Pa / cfm ² ).

ρ is the air density, A is the tile area, and K is the loss factor. The pressure difference and the air flow are related as follows.
Δρ = R · f ² _airflow
The pressure difference is re-measured with a real-time sensor and Δρ = R · f ² _airflow is used to calculate the airflow through each tile. As mentioned above, airflow is applied as a boundary.

  Once the model is defined and set up (ie, a graphical display is provided, all domains are defined, etc.), it can be based on user input or (eg, as described above, airflow changes or domains Temperature zone, either triggered by changes measured in the data center using a sensor network (which can detect the addition / removal of sources / sinks, or both) Can be modeled (as described above). FIG. 10 is a diagram illustrating the visualization of temperature zones on the interface 800 (of FIG. 8). In the exemplary embodiment shown in FIG. 10, gray shading is used to differentiate the energy efficiency of the various temperature zones, for example, the shading used within temperature zone 1002 shows the highest efficiency. As is apparent above, efficiency can be characterized by “utilization” or more desirably “COP”. Increasing “removal heat” at a given nominator COP increases utilization, and thus efficiency (ie utilization (UT) = removal heat / nominal heat load removal capability) or coefficient of performance (COP) (COP = removed heat / power consumption for ACU fan) increases. Next, the bar graph presented at the bottom of the interface 800 shows utilization and COP. Lines 1004 and 1006 on the bar graph correspond to the ACU discharge and return temperatures, respectively, from static data (which can be edited with the MMT client) or from the data center (eg, as described above). Can be obtained either from real-time data (sent from the real-time sensor into the MMT client) to update the conditions in For illustrative purposes, one of the temperature zones (with a utilization of 54% and 9 COPs) is shown linked to the corresponding bar graph by arrow 1008.

が取得されると、温度分布に関して解決するために、これがエネルギー式

で使用され、温度は（たとえばサーバの吸気口および排気口での）境界として指示される）。したがって、温度分布は（たとえば２次元の場合）、ｘおよびｙ座標の関数Ｔ（ｘ，ｙ）として計算可能である。穿孔タイルがどこにあるか（ｘｔ，ｙｔ）を知ることによって、タイル排出温度Ｔｄ＝Ｔ（ｘｔ，ｙｔ）を取得することができる。 Finally, it should be noted that the technique can also be used to determine the respective cooling capacity from each vent or perforated tile. That is, (as described in Hamann) in relation to the temperature model, discharge temperature T _D of the vent hole or perforation tiles can be calculated (i.e., velocity field

Is obtained, this is the energy formula to solve for the temperature distribution

And the temperature is indicated as a boundary (eg at the server inlet and outlet). Thus, the temperature distribution (eg in the case of two dimensions) can be calculated as a function T (x, y) of x and y coordinates. By knowing where the perforated tile is (xt, yt), the tile discharge temperature Td = T (xt, yt) can be obtained.

によって、各通気孔または穿孔タイルについて、１タイル当たりの冷却力を決定することができる。速度ベクトルが（本実施形態におけるように、境界として測定および使用されるのではなく）計算される場合、流れは、

によって、タイルまたは通気孔の領域全体にわたって垂直速度ベクトル成分（ほとんどの場合ｖ_ｚ）を積分することによって、取得可能である。この機能の例が図１１に示されている。具体的に言えば、図１１は、データ・センタ内の冷却能力を示す概略図１１００である。図１１では、レイアウト表示が擬似３次元イメージとして３次元棒グラフとオーバレイされており、棒の高さは冷却容量（すなわち冷却力）を示す。 Combining the airflow velocity or total airflow with the allowable _inlet temperature T _{inlet of} the server,

Allows the cooling power per tile to be determined for each vent or perforated tile. If the velocity vector is calculated (rather than measured and used as a boundary as in this embodiment), the flow is

Can be obtained by integrating the vertical velocity vector component (in most cases v _z ) over the entire area of the tile or vent. An example of this function is shown in FIG. Specifically, FIG. 11 is a schematic diagram 1100 illustrating the cooling capacity within a data center. In FIG. 11, the layout display is overlaid with a three-dimensional bar graph as a pseudo three-dimensional image, and the height of the bar indicates the cooling capacity (ie, cooling power).

  FIG. 12 is a diagram illustrating an example method 1200 for tracing an airflow velocity vector to an airflow source such as an ACU. Method 1200 represents an exemplary process for performing step 214 of method 200 described above (of FIG. 2). In step 1202, the process begins at a given location (which is also considered the initial location if this is the first iteration of the process) and a given (eg, initial) location is selected. According to an exemplary embodiment, if the space being modeled is a data center, each location described according to this process can be any given tile (not only perforated tiles, but any Corresponds to the position of the tile. In that case, the initial position can be any tile. The process starts at that initial tile (position) and can then loop through all tiles in the data center. In step 1204, a velocity vector for a given position (the initial position in the first iteration of the process) is referenced (ie, the velocity vector across the data center has already been determined, eg, step 200 of the method 200 described above. 212). As described above, the air flow velocity vector across the space, ie the data center, is determined using finite element techniques (ie within multiple subdomains) based on the air flow data obtained from the source / sink. can do. However, if the velocity vector for a given position has not yet been determined, it can be extracted using standard interpolation techniques. For example, when the flow path is calculated, the air flow vector finally reaches a position where it has not been specifically calculated yet. However, of course, there is a position close to the position where the airflow vector is calculated. Several different approaches have been tested for these scenarios. One approach is to use the closest velocity vector. Another approach is to use a distance weighted average of multiple (eg 9) adjacent velocity vectors. In general, however, any interpolation technique can be used.

In step 1206, a new next position is determined or calculated. In this step, the next tile is selected using the above example with the data center tile as the location point. In the two-dimensional domain, the next positions xt and yt can be selected as follows: xt (steps) = (xt (steps−1) + vx ^* stepsize ^* n) and yt (steps) = (yt (steps− 1) + vy ^* stepsize ^* n), where the velocity vector magnitude is v = sqrt (vx ² + vy ² ) and n = 0.2 / v. v = sqrt (vx ² + vy ² ). . . n = 0.2 / v is a parameter that controls the step size while associating with vx and vy and prevents the step from becoming too small when the velocity vector is small. In other words, if the speed is very slow, the variable n is introduced to ensure that the resizable step still moves from one position to the next so that the method is not too slow. The The step size can be selected by the user.

  In step 1208, it is determined whether the new location is in the area of the airflow source, eg, ACU. If the flow path intersects the ACU, at step 1210, the ACU is assigned a previous position. For example, if tile 1 is the initial (previous) position and tile 2 is the new position, tile 2 is in the area of ACU1 and tile 1 is assigned to ACU1. Next, the process (space, eg, data center, etc.) is made by referring to the velocity vector for tile 2 (eg, step 1204), eg, determining another new location of tile 3 (step 1206), etc. Iterate over n positions, eg tiles, throughout. The ACU position is recognized from the graphic display of the MMT client, for example. The display includes the ACU's x, y coordinates, as well as the ACU's width, length, and height (ie, region definition). If xt and yt (see step 1206) are within the ACU domain, then another new position is determined.

  On the other hand, if the new location (eg, tile 2 using the above example) is not in the region of the ACU, at step 1212 it may be possible to make a determination as to whether too many steps have been created, That is, if the flow path does not end with an ACU, there is a maximum number of steps. Another way to consider this is that a limit may be imposed on the number of times steps 1204 and 1206 can be repeated without having a location (eg, a tile) within the region of the airflow source. In fact, if too many steps have been created (ie, the limit has been reached), at step 1214, the previous location, eg, tile 1, means that no temperature zone has been assigned to that location, “no zone” Is specified. The process then refers to the new position, eg, the velocity vector for tile 2 for n positions throughout the space (step 1204), eg, determines another new position for tile 3 (step 1206), And so on. On the other hand, if the maximum number of steps has not been reached (ie, the limit has not been reached) (and the new location, eg, tile 2 is not in the region of the ACU (step 1208)), the process starting at step 1204 is That is, iterating by referring to a new location, eg, the velocity vector for tile 2 (step 1204), eg, determining another new location of tile 3 (step 1206), and so on. The method 1200 is repeated for each of the n locations or tiles in the space or data center. In this way, each velocity vector is traced to a particular airflow source (eg, ACU) or designated as not associated with a particular temperature zone. Exemplary code for tracing a velocity vector to a particular airflow source and thereby defining a temperature zone within the space is provided below.

  Turning now to FIG. 13, a block diagram of an apparatus 1300 for modeling a temperature zone in a space, such as a room, such as a data center 100 (FIG. 1), according to one embodiment of the present invention is shown. Has been. It should be understood that apparatus 1300 represents one embodiment for performing method 200 of FIG.

  Apparatus 1300 includes a computer system 1310 and removable media 1350. The computer system 1310 includes a processor device 1320, a network interface 1325, a memory 1330, a media interface 1335, and an optional display 1340 (eg, for displaying the graphical interface 800 of FIG. 8 described above). Network interface 1325 enables computer system 1310 to connect to a network, and media interface 1335 enables computer system 1310 to interact with media such as a hard drive or removable media 1350. To do.

As is known in the art, the methods and apparatus discussed herein comprise a machine-readable medium containing one or more programs that, when executed, implement embodiments of the present invention. It can be distributed as a product itself. For example, machine-readable media, provides a graphical overview of the space, fractionated constant at least one domain in the space for modeling, subdivided domain to a set of interconnect to salicylate subdomains multiple nodes To create a mesh in the domain, identify airflow sources and sinks in the domain, obtain airflow measurements from one or more of the airflow sources and sinks, and And the airflow measurements taken from the sinks are used to determine the airflow velocity vector at the center of each subdomain, tracing each velocity vector to one of the airflow sources, Includes a program where the combination of traces to a given one of them represents a temperature zone in the space It can be.

  The machine readable medium can be a recordable medium (eg, a flexible disk, a hard drive, an optical disk such as removable media 1350, or a memory card) or a transmission medium (eg, fiber optic, world wide, etc.). Web, cable, or network with radio channels using time division multiple access, code division multiple access, or other radio frequency channels. Any medium known or developed that can store information suitable for use with a computer system may be used.

  The processor device 1320 can be configured to implement the methods, steps, and functions disclosed herein. Memory 1330 can be distributed or local, and processor 1320 can be distributed or single. Memory 1330 can be implemented as electrical, magnetic, or optical memory, or any combination of these or other types of storage devices. Further, the term “memory” is sufficient to encompass any information that can be read from or written to an address in the addressable space accessed by the processor device 1320. Should be interpreted broadly. With this definition, information on the network accessible via the network interface 1325 will still be in the memory 1330 because the processor device 1320 can retrieve the information from the network. Note that each distributed processor comprising processor device 1320 generally includes its own addressable memory space. Note also that some or all of computer system 1310 can be incorporated into an application specific integrated circuit or a general purpose integrated circuit.

  Optional video display 1340 is any type of video display suitable for interacting with a human user of device 1300. In general, video display 1340 is a computer monitor or other similar video display.

The following is exemplary code for the PV-WAVE (R) programming language to trace a velocity vector to a particular airflow source, thereby defining a temperature zone within the space.

  As mentioned above, although exemplary embodiments of the present invention have been described in the present specification, the present invention is not limited to these precise embodiments, and without departing from the scope of the present invention. It will be appreciated that various other changes and modifications can be made by the vendor.

Claims (17)

A method for modeling a temperature zone in a space given by a streamline of airflow , wherein the device comprises:
Providing a graphical representation of the space, the graphical representation comprising a layout of a computer equipment rack, an air conditioning unit, and vent or stenosis tiles ;
Defining at least one region within the space for modeling;
Creating a mesh in the region by subdividing the region into a set of subdivided regions;
Identify multiple locations within the region where airflow enters the region (hereinafter referred to as multiple airflow inflow locations) and multiple locations where airflow exits the region (hereinafter referred to as multiple airflow outflow locations) And steps to
Obtaining an airflow measurement from one or more of the plurality of airflow inlet locations and the plurality of airflow outlet locations;
Using the airflow measurements obtained from the one or more of the plurality of airflow inflow locations and the plurality of airflow outflow locations, a central airflow velocity vector for each subdivision region is determined. And steps to
Tracing each of the air flow velocity vectors to one of the air flow inlet locations, wherein the air flow velocity vectors collectively represent air flow streamlines , and the temperature zone is the trace. in combination, provided by the streamlines of the airflow, the is defined by the space, seen including to perform the steps of said trace,
The step of tracing comprises:
Selecting a given position in the space;
Referencing the airflow velocity vector determined at the center of the subdivision region at the selected given location;
Determining a next position calculated from the selected given position within the space, wherein the next position is determined from the given given position and the airflow velocity vector; Calculating said determining step;
Referencing the airflow velocity vector and determining the next position for each of the n positions throughout the space, and repeating the steps based on the next position during the iteration A next position is determined, and if the next position is in a region where a given one of the airflow inflow locations is in an area, the selected given position is determined to be the given one airflow inflow. Assigning to a location, said repeating step and
The method further comprising:   The method of claim 1, wherein the space comprises a room.   The room is configured to take hot air from the computer equipment rack and the computer equipment rack and exhaust cooling air delivered to the computer equipment rack through a plurality of perforated tiles in a raised floor into the underlying floor plenum. 3. A method according to claim 2, comprising a raised floor cooling system comprising one or more computer air conditioning units.   The method of claim 3, wherein the region is defined by the dimensions of the foundation floor plenum.   5. A method according to any one of the preceding claims, wherein the airflow measurements from the airflow inlet location and the airflow outlet location are obtained using a moving measurement technique.   5. The method of claim 4, wherein the perforated tile comprises the air flow outflow location and the air conditioning unit comprises the air flow inflow location. The device is
Tracing each of the velocity vectors to one of the air conditioning units, wherein the combination of traces to a given one of the air conditioning units represents a temperature zone in the space; The method of claim 6, further comprising performing the tracing step. The device is
Repeating the step of identifying, obtaining, using, and tracing in response to a change in airflow at one or more of the airflow inflow location and the airflow outflow location in the region. The method according to claim 1, comprising performing. The device is
Repeating the step of identifying, obtaining, determining, and tracing in response to removal of one or more of the airflow entry location and the airflow exit location from the region. The method according to claim 1, comprising: The device is
Further comprising repeating the steps of identifying, obtaining, determining, and tracing in response to one or more new airflow inflow locations and airflow outflow locations being added to the region. The method according to any one of claims 1 to 9. The device is
The method according to claim 1, further comprising: determining a cooling capacity for each airflow inlet location. The device is
The method further comprises the step of periodically repeating the steps of obtaining, determining, and tracing to update the airflow measurements from the airflow inflow location and the airflow outflow location. The method as described in any one of -11. Without having a position that is within the area where the air flow outlet location of a given position wherein the selected further comprises the step of providing a limit to the number of times the steps are performed to the iteration of claim 1 Method. 14. The method of claim 13 , further comprising determining whether the iterative step has been performed too often whenever the next location is not within the region where the airflow outflow location is. . 14. The method of claim 13 , further comprising designating the selected given location as not associated with the temperature zone whenever the limit is reached. A program for modeling a temperature zone in a space, given by streamlines of air flow, causing an apparatus to perform the steps of the method according to any one of claims 1 to 15. program. An apparatus for modeling a temperature zone in a space given by a streamline of airflow ,
Memory,
At least one processor device coupled to said memory, comprising:
16. The apparatus comprising: the at least one processor device capable of performing the steps of the method according to any one of claims 1-15 .

Images